WorksPhysics and Astronomy
The Ionospheric Cut-off as a Potential Radio Technosignature
Abstract
Radio emission generated on a planetary surface must cross the planet's ionosphere before escaping into space: below the local plasma frequency it is reflected back towards the surface. We investigate this process for broadband radio technosignatures (BRaTs) using a Chapman-layer description of photoionised atmospheres. The plasma frequency depends only weakly on stellar XUV irradiation, scaling as , implying turnover frequencies of 5-40 MHz for most terrestrial habitable-zone planets, potentially rising towards 100 MHz in extreme environments. The basic signature is simple: the continuum of a surface-origin BRaT should be strongly suppressed below a characteristic frequency that can be estimated from the XUV output of the host star, at every orbital phase and for any viewing geometry. Oblique propagation, magnetic fields, and collisional absorption broaden the cut-off into a smooth escape function, , where is the orbital phase. Because the day-night ionisation pattern is anchored to the substellar point rather than to the rotating surface, disc integration adds a second, more discriminating signature: the half-escape frequency varies predictably (by about 40-50 per cent over the orbit, depending on the angular emission pattern of the leakage) and remains locked to the orbital ephemeris. Finite atmospheric response times introduce modest phase lags but preserve the modulation. A low-frequency turnover alone is not unique to technology, but one that varies periodically with the orbital period would be difficult to reproduce through known astrophysical processes. Although observationally challenging, the ionospheric cut-off is a potential technosignature diagnostic and a novel probe of exoplanet ionospheres.
Introduction
For over sixty years, radio SETI has prioritised narrowband, largely monochromatic emissions, a strategy driven by their energetic efficiency and the absence of natural confounding sources. While this approach remains highly relevant (Cocconi & Morrison 1959; Tarter 2001; Enriquez et al. 2017; Price et al. 2020), the parameter space for potential technosignatures is significantly broader. Recent work has begun to explore the complementary regime of broadband radio technosignatures (BRaTs), in which the integrated, unintentional electromagnetic leakage of a planetary-scale technological society produces a measurable enhancement in the radio continuum of an unresolved exoplanet (Garrett 2026). BRaTs offer a new and largely unexplored search domain.
A central question for any BRaTs search programme is the choice of observing frequency. On Earth, the ground-accessible radio window is bounded at low frequencies by the terrestrial ionosphere, near MHz, and at high frequencies by increasing atmospheric and molecular absorption beyond 50 GHz Wilson et al. (2013). The same physics, with very different parameters, governs the escape of artificial radio emission from the surface of an exoplanet into space. Below the local plasma frequency, the electromagnetic wave is reflected by the ionosphere and ultimately dissipated within the atmosphere. Well above the local plasma frequency, the wave escapes into space with little impedance. As emphasised by Garrett (2026), the implication for SETI can be highly significant: low-frequency leakage from worlds whose ionospheres are appreciably denser than Earth’s may be entirely invisible to a distant observer.
This paper expands on the possible impact of the ionospheric cut-off on BRaTs as noted by Garrett (2026). It develops the ionospheric cut-off as both a constraint and a possible signature for BRaTs. Our motivation is threefold. First, the physical conditions governing the ionospheric cut-off on exoplanets remain poorly explored, especially for terrestrial worlds in the habitable zones of low-mass, active stars. Secondly, the cut-off is itself a simple and testable prediction: radio emission that originates at a planetary surface cannot be seen below the plasma frequency of the overlying ionosphere, so the continuum spectrum of a BRaT should turn over at a frequency that can be estimated in advance from the XUV output of the host star. This holds at every orbital phase, and testing it requires no knowledge of the orbit. Finally, we argue that because the cut-off frequency itself varies in a deterministic, geometry-locked manner, it offers a diagnostic that known natural sources are unlikely to mimic in full. Natural processes can produce low-frequency spectral turnovers, and some, notably auroral cyclotron-maser emission and star–planet interactions, can produce periodic variability (Section 5.5). However, no known natural process presents the specific combination that the ionospheric cut-off predicts: a smooth, weakly polarised turnover whose characteristic frequency varies with planetary phase. The detection of a broadband planetary radio continuum exhibiting a low-frequency turnover, whose frequency oscillates in phase with the planet’s orbital geometry, would therefore provide strong supporting evidence for a planetary-surface broadband radio technosignature.
The paper is organised as follows. § 2 summarises the plasma physics of the ionospheric cut-off, including the effects of the planetary magnetic field and oblique propagation. § 3 reviews ionospheric properties across the Solar System as empirical benchmarks, and § 4 builds on this with models of the expected cut-off frequencies of terrestrial exoplanet ionospheres as a function of their stellar host parameters. § 5 develops the central result: the orbital-phase modulation of the apparent ionospheric frequency cut-off, which an unresolved external observer would measure, and the natural processes that could mimic it. § 6 outlines an observational strategy and the criteria a candidate detection should satisfy. § 7 discusses the broader value of the cut-off as a planetary atmospheric diagnostic, and we draw our conclusions in § 8.
Ionospheric plasma physics
The plasma frequency
Consider first an idealised ionosphere: a cold, unmagnetised plasma with a smooth electron-density profile that varies only with altitude, free of horizontal gradients or irregularities (collisions and magnetic fields are introduced below). The response of the free electrons to an externally applied electromagnetic wave gives rise to the plasma frequency (Rybicki & Lightman 1986; Kelley 2009)
where is the elementary charge, the electron mass, and the vacuum permittivity. Electromagnetic waves with angular frequency have an imaginary wavenumber and are reflected. Waves with propagate with a refractive index
approaching unity at high frequencies.
For an inhomogeneous ionosphere stratified primarily in altitude, radio-wave propagation is determined by the electron-density profile rather than by altitude itself. Altitude matters only because the electron density, and hence the local plasma frequency, vary with height. A wave of frequency can propagate only through regions where . Reflection occurs at the turning point where the local plasma frequency first becomes equal to the wave frequency, . The maximum electron density encountered along the propagation path therefore determines whether a wave can escape into space. We refer to the corresponding frequency as the critical frequency , which for vertical incidence is equal to the plasma frequency at the density peak of the ionospheric layer.
Oblique propagation
A surface-based transmitter emits radiation into the upward hemisphere, spanning a range of propagation directions. While vertically propagating waves encounter the minimum ionospheric column density, obliquely propagating rays traverse a longer path through the stratified plasma.
For vertical propagation, electromagnetic waves escape provided their frequency exceeds the local ionospheric cut-off, . For a ray propagating at an angle to the local vertical, the tangential component of the wave vector is conserved in a horizontally stratified medium; it is the normal component that determines whether the ray continues upward or reaches a turning point. In the plane-parallel approximation, the component of the incident wave vector normal to the ionospheric layers is reduced by a factor , yielding the escape criterion
A consequence of this secant-law dependence is that emission at frequencies modestly above can still be trapped if launched at sufficiently low elevation. For a planetary-scale leakage source, the limb-grazing components are systematically suppressed by the obliquity factor, broadening the spectral turnover. The broadening is of order unity in for a broad, quasi-isotropic distribution of launch angles, although the precise factor depends on the transmitters’ angular radiation patterns, on ionospheric curvature and refraction, and on the quantity being observed. The ionospheric cut-off is therefore smoother than an abrupt break at a single, well-defined frequency.
Consequently, the minimum frequency required for escape increases with the propagation angle. Rays directed close to the horizon () require progressively higher frequencies to penetrate the ionosphere, while exactly horizontal propagation does not lead to escape within the plane-parallel approximation. This formal divergence is an artefact of the plane-parallel approximation: for a real spherical planet the ionosphere is curved, so a ray that is initially horizontal with respect to the surface does not remain parallel to every overlying layer, and its actual trajectory is determined by planetary curvature, the finite thickness of the ionosphere, and refraction. A simple regularisation, which we use in § 5.2, treats the ionosphere as a thin spherical shell at height above a planet of radius . A ray leaving the surface at zenith angle then crosses the shell at the smaller angle , with , so that the obliquity factor becomes
which reproduces the secant law for small but remains finite at the limb, for (Earth’s F2 peak). Refraction and the finite thickness of real layers modify this further, but the thin-shell form captures the essential regularisation.
Magnetised plasma effects
So far, we have treated the ionosphere as an unmagnetised plasma. This is a useful first approximation, but real planets often have magnetic fields. In a magnetic field, free electrons do not simply oscillate back and forth in response to a radio wave. They also spiral around the magnetic field lines. This changes the way radio waves propagate through the plasma (Budden 1985).
The characteristic frequency of this spiral motion is called the electron gyrofrequency:
Here is the magnetic-field strength, is the electron charge, and is the electron mass. For Earth, , so
This frequency is significantly lower than the daytime ionospheric plasma cut-off, which can approach approximately 10 MHz (the peak plasma frequency varies considerably with location, local time, season, and solar activity), but it is not entirely negligible.
The magnetic field splits the radio-wave propagation into two natural modes, usually called the ordinary mode (O-mode) and the extraordinary mode (X-mode). These modes behave slightly differently because the electrons are forced to move in circular paths around the magnetic field. For waves travelling perpendicular to the magnetic field, the upper X-mode cut-off frequency is
If the plasma frequency is much larger than the gyrofrequency, , this becomes approximately
This is usually a good approximation for Earth-like ionospheres. For propagation parallel to the field the natural modes are instead the right- and left-hand circularly polarised (R and L) modes; equation (5) then gives the R-mode cut-off, while the L-mode cut-off,
lies below .
The main point is that a magnetic field splits the cut-off rather than simply shifting it: the X-mode cut-off lies above , the O-mode cut-off remains near , and other mode boundaries can lie below it. Instead of a single cutoff frequency, the ionosphere has different cut-offs for different wave modes and directions of travel. The exact answer depends on the angle between the wave direction and the magnetic field, as well as the local plasma density and magnetic field strength. Again, the overall effect is to smear out the spectral turnover rather than to produce a perfectly clean edge.
Electron–neutral collisions also provide an additional source of attenuation near the plasma cut-off ((Kelley 2009)). Collisions do not alter the formal plasma cut-off at ; rather, they introduce absorption that broadens the transition, so that appreciable power may escape only at frequencies somewhat above the nominal cut-off. This effect is strongest in dense, weakly ionised ionospheres or lower atmospheric plasma layers, where the neutral density, and hence the electron–neutral collision frequency, is high. A large electron density alone does not produce strong collisional absorption.
The Chapman layer and time variability
The ionosphere is produced when ultraviolet and X-ray radiation from a star ionises the upper atmosphere. The electron density is therefore not uniform with altitude. At very high altitudes, there are relatively few neutral atoms available to ionise, while at lower altitudes, much of the ionising radiation has already been absorbed. As a result, the electron density reaches a maximum at an intermediate height.
A simple and widely used description of this structure is the Chapman layer ((Chapman 1931; Kelley 2009)). In this model, the equilibrium electron density at altitude is
where is the altitude in units of the atmospheric scale height , and and in this expression are the height and density of the peak for an overhead star (). At zenith angle , the peak rises by and its density falls as , consistent with the peak scaling derived below. The angle is the stellar zenith angle, i.e. the angle between the local vertical and the direction to the host star. At local noon , while near sunrise and sunset approaches .
Although equation (6) describes the detailed vertical structure of the ionosphere, the quantity most relevant for radio-wave escape is the peak electron density, , since the plasma cut-off is determined by the densest part of the layer.
The scaling of can be understood from a simple ionisation–recombination balance (the -Chapman equilibrium; e.g. Chapman 1931; Schunk & Nagy 2009). In equilibrium, the rate at which ionising photons create free electrons must be balanced by the rate at which electrons and ions recombine. If the ionisation rate is proportional to the incident photon flux, , and the recombination rate is proportional to , then
Here is the ionising photon flux at the top of the atmosphere and is the recombination coefficient of the dominant ion species. Solving for the equilibrium electron density gives
Since the plasma frequency scales as , the maximum plasma frequency in the ionosphere follows
Equations (7)–(9) are proportionality relations: dimensionless atmospheric and geometric factors – the neutral scale height , the ionisation efficiency, and the Chapman-function geometry – have been absorbed into the constants of proportionality, or equivalently into an effective . They therefore describe scalings among atmospheres of broadly similar structure; strong differences in scale height or composition enter through these omitted factors, a point we return to in § 4.
This quarter-power dependence is an important scaling factor that plays a central role in the discussion of exoplanet ionospheres (§ 4) and the phase modulation of the cut-off frequency (§ 5). Because the plasma frequency depends only on the fourth root of the ionising flux, even large changes in stellar irradiation produce relatively modest changes in the cut-off frequency. For example, doubling the cut-off frequency requires a sixteen-fold increase in the ionising flux. Similarly, the variation in across the dayside causes the ionospheric electron density, and hence the plasma cut-off frequency, to vary predictably between local noon and the day–night terminator. However, changes in the cut-off frequency are only a factor of a few between local noon and the day–night terminator, rather than orders-of-magnitude variations.
Ionospheres of the Solar System
The Solar System provides empirical anchors that span more than three orders of magnitude in peak electron density and provide empirical anchors for much of the parameter space relevant to BRaTs (Table 1). Earth’s mid-latitude F2 layer reaches at solar maximum and during the day, dropping by an order of magnitude at night and during solar minimum ((Kelley 2009)). Venus, despite receiving roughly twice the solar extreme-ultraviolet (EUV) flux of Earth (inverse-square scaling at 0.72 au), has a comparable cut-off because its dense CO atmosphere recombines more efficiently - it supports a dayside ionosphere with near 140 km altitude ((Brace et al. 1979; Bauer & Lammer 2004)). Mars exhibits a
| Body | (m) | (MHz) | Ref. |
| Mercury | a | ||
| Venus (day) | 7 | b | |
| Earth (day, max) | – | 9–13 | c |
| Earth (night) | – | 3–5 | c |
| Mars (day) | 4 | d | |
| Jupiter | 4 | e | |
| Saturn | – | 1–3 | e |
| Titan | 0.5 | f | |
| Uranus | 1 | e | |
| Neptune | 1 | e |
Table 1. Representative ionospheric peak electron densities and corresponding vertical plasma cut-off frequencies in the Solar System.
References: (a) Killen et al. (2007); (b) Brace et al. (1979); (c) Kelley (2009); (d) Withers (2009); (e) Yelle & Miller (2004); (f) Wahlund et al. (2005).
tenuous ionosphere with at 130 km (Withers 2009).
Several features of Table 1 are worth noting. First, the variation among bodies, despite differences of orders of magnitude in stellar irradiation, atmospheric composition, and magnetic environment, spans a factor of in , from Mercury to Earth’s daytime maximum – equivalent to a factor of in . This range, modest compared with the spread in solar irradiation alone across the same bodies, partly reflects the quarter-power scaling of equation (9) and partly the fact that, in steady state, recombination and ionisation balance limit excursions in . Secondly, the night-side of a predominantly photoionised ionosphere typically exhibits a significantly lower cut-off than the day side, by a factor of typically a few, providing the empirical basis for the day–night modulation argument developed later. This contrast is a common tendency rather than a universal rule: it depends on recombination, dayside-to-nightside transport, particle precipitation, auroral processes, and magnetospheric plasma sources, any of which can sustain a substantial nightside density. Thirdly, despite their large sizes, the giant planets have only weakly ionised upper atmospheres: relatively few free electrons are produced by solar EUV radiation at these distances, and those that are produced are rapidly removed by dissociative recombination with molecular ions such as , a process that neutralises the ion and yields neutral fragments.
Exoplanet ionospheric cut-offs
The discussion of § 2 establishes that the radio cut-off of a planetary ionosphere is set by the peak electron density at the layer maximum, which, in turn, is set by the balance between stellar photoionisation and ionic recombination (equation (9)). To predict the cut-off frequency for a given exoplanet, we need to know how much ionising radiation the planet receives at the top of its atmosphere and how efficiently its dominant ions recombine.
Stellar EUV and X-ray driving
The photons responsible for ionising planetary atmospheres lie almost entirely in the extreme-ultraviolet and X-ray bands (XUV). We adopt the conventional integration boundary (), which corresponds to the ionisation threshold of atomic hydrogen; it is a bookkeeping convention rather than a universal photoionisation threshold, since several molecular constituents ionise at lower energies (e.g. for and for ), so that the relevant band is composition-dependent. At these energies, the photons can liberate electrons from the dominant neutral constituents (, , , ), and the strength of the XUV flux at the top of the atmosphere directly sets the primary ionisation rate and hence the ionospheric electron density (France et al. 2013; Peacock et al. 2019).
Two features of low-mass stars matter here. First, even in their quiescent state, M-dwarfs emit a larger fraction of their bolometric output in the XUV than Sun-like stars do, often by several orders of magnitude in the most active cases. Second, the habitable zones of M-dwarfs lie much closer to the host star than the Sun’s habitable zone, mainly because M-dwarfs have low bolometric luminosities. Both effects compound, so a habitable-zone planet orbiting an M-dwarf receives XUV flux many times that of Earth’s.
A convenient parameterisation is the photon flux at the top of the planetary atmosphere,
where is the stellar XUV luminosity, is the orbital separation, and is the mean photon energy in the relevant band. For Earth, (e.g. Schunk & Nagy 2009). Proxima Centauri b (Anglada-Escudé et al. 2016), in the habitable zone of a moderately active M5.5V star, receives an XUV irradiance times that of Earth (Ribas et al. 2017); assuming a mean XUV photon energy comparable to the solar value, this corresponds to . The early M-dwarf AD Leo (M3V) places its habitable zone in a still harsher XUV environment, with (estimated from the quiescent XUV spectrum adopted by Segura et al. 2010). The seven known terrestrial planets of TRAPPIST-1 (Gillon et al. 2017) span a more modest –15 in quiescence, although the system is also known for strong, frequent flares (Wheatley et al. 2017; Vida et al. 2017).
Scaling the cut-off frequency to other planets
Combining the Chapman scaling of equation (9) with the plasma-frequency relation of equation (1), and normalising to Earth’s daytime peak ( at ), we obtain
Equation (11) implicitly assumes that the atmospheric scale height, ionisation efficiency, illumination geometry, and XUV spectral shape are Earth-like (cf. § 2); it should be read as a scaling among broadly terrestrial atmospheres rather than as a general formula. Here is a representative laboratory dissociative-recombination coefficient for molecular ions and at (Schunk & Nagy 2009), adopted as the loss coefficient in the closure of equation (7). We emphasise what this coefficient does and does not represent. The terrestrial F2-region peak is dominated by atomic , which is removed through ion–neutral reactions that produce molecular ions ( and ); these molecular ions then undergo rapid dissociative recombination (molecular ions can replace as the dominant F-region ion under particular disturbed or high-latitude conditions; e.g. Banks et al. 1974). The first, ion–neutral step is generally rate-limiting and is linear rather than quadratic in , so a phenomenological coefficient that reproduced the combined sequence would depend on neutral density, composition, altitude, temperature, and itself, and would not equal the molecular-ion dissociative-recombination coefficient. We therefore keep the two quantities distinct: the ratio in equation (11) compares laboratory dissociative-recombination coefficients of the
dominant molecular ions across atmospheric chemistries – a defined, if simplified, comparison – whereas the empirical atmospheric relaxation time invoked in § 4.4 and § 5.3 is a separate parameter, introduced and calibrated there. A CO-dominated atmosphere recombines through O with ; an H-dominated atmosphere (e.g. a retained primordial envelope on a sub-Neptune) recombines mainly through H with –, depending on electron temperature and the degree of internal excitation. Because the dependence is fourth-root, the cut-off is only weakly sensitive to uncertainty in a fixed effective recombination coefficient, although substantially different atmospheric chemistries may also alter the scale height and the dominant loss processes.
Representative cases
Table 2 applies equation (11) to a small set of representative habitable-zone targets. For an Earth-analogue at 1 AU around a Sun-like star (G2V), we recover MHz by construction. The quiescent late-G dwarf Ceti (G8V) hosts the candidate planet Ceti e in a slightly weaker XUV environment than Earth, with MHz. A hypothetical terrestrial planet in the habitable zone of the young, active K-dwarf Eridani would receive , giving MHz (the known planet Eri b is a giant on a near-circular orbit at au (Thompson et al. 2025) and is not itself a BRaTs target). For terrestrial planets around active M-dwarfs, the cut-off climbs into the tens of MHz: Proxima b sits at MHz, and a planet in AD Leo’s habitable zone at MHz.
| Target | quiet (MHz) | XUV enh. (factor) | flare (MHz) | |
| Sun–Earth | 1 | 10 | 3 | 13 |
| Ceti e (G8V) | 0.6 | 9 | 3 | 12 |
| Eri HZ (K2V) | 15 | 20 | 10 | 36 |
| TRAPPIST-1e (M8V) | 2 | 12 | 100 | 38 |
| Proxima b (M5.5V) | 60 | 28 | 100 | 89 |
| AD Leo HZ (M3V) | 200 | 38 | 100 | 120 |
Table 2. Table 2. Estimated quiescent and flare-state plasma cut-off frequencies for representative habitable-zone targets, computed from equation (11) with Earth-like atmospheric composition. The flare column applies equation (11) with the target-dependent sustained XUV enhancement factor listed in the fourth column: a modest for the Sun-like and late-G hosts, for the active K dwarf, and – a fiducial strong-flare scenario – for the M dwarfs. Because , even a enhancement raises by only a factor of . The flare values are equilibrium responses, attained only if the enhancement is sustained for longer than the ionospheric response time (see § 4.4).
A few general remarks follow from Table 2. For terrestrial planets in the habitable zones of Sun-like and K-dwarf hosts, the cut-off remains close to Earth’s value, in the 5–30 MHz range; low-frequency BRaTs leakage from such systems is in principle accessible to a sufficiently sensitive metre-wave observatory. For active M-dwarfs, the cut-off climbs into the tens of MHz, potentially rising towards MHz in the most extreme flare-driven or strongly irradiated environments, which significantly restricts the accessible window below 1 GHz but does not close it. In all cases, observing at GHz remains a safe choice for detection-driven BRaTs searches that are willing to forego the diagnostic value of the cut-off itself (§ 5).
Flares and the recombination floor
The numbers in the third column of Table 2 refer to the quiescent state of each star. M-dwarfs, however, are well known to produce frequent, energetic flares with optical energies typically – erg ((Howard et al. 2019)). The conversion from optical flare energy to instantaneous XUV enhancement is uncertain and flare-dependent, but enhancements of factors – above the quiescent XUV level have been inferred for strong events, on timescales of minutes to hours; we therefore treat the enhancement factor as a free parameter in what follows. Through equation (11) a XUV enhancement raises by a factor of ; a enhancement by . The flare column of Table 2 applies target-dependent sustained enhancements – for the Sun-like and late-G hosts, for the active K dwarf, and the fiducial strong-flare scenario of for the M dwarfs – giving transient cut-offs of MHz for Proxima b and MHz for a planet in AD Leo’s habitable zone. Reaching the gigahertz regime through flaring is effectively impossible: because , raising the cut-off from 28 MHz (Proxima b) or 38 MHz (AD Leo) to 1 GHz would require XUV enhancements of or relative to quiescence, far beyond the magnitude of any observed flare.
These flare-state values are equilibrium responses: they assume the ionosphere has time to come into a new Chapman equilibrium with the elevated XUV flux. In practice, the time required is set by the recombination timescale. For a simple ionosphere dominated by one positive ion species, electron loss occurs through recombination,
so the loss rate is proportional to the product of the electron and ion number densities,
where is the effective recombination coefficient. Since the plasma is quasi-neutral on the relevant scales, , and hence
which is the loss term in equation (7). A characteristic recombination time is then obtained by dividing the electron density by its loss rate (e.g. (Bauer & Lammer 2004; Schunk & Nagy 2009)),
evaluated at the peak of the ionospheric layer. For the daytime Earth-calibrated values adopted here ( MHz, i.e. , with ), equation (12) evaluates to s, rising to tens of seconds at typical night-side densities; it should be regarded as a lower bound on the true response time. In Earth’s F region the effective relaxation time is far longer – tens of minutes to hours – because the peak is dominated by atomic O, whose removal is limited by the slower ion–neutral chemistry described in § 4, and because transport contributes to the balance; in that regime the simple model is no longer sufficient. Whether a flare drives the cut-off to its equilibrium value therefore depends on which response time applies: bursts shorter than the relevant response time produce only partial enhancements. The flare entries in Table 2 should accordingly be read as values attained only in sustained events: a brief flare burst leaves the ionosphere largely undisturbed, while an elevated XUV state lasting longer than the atmospheric response time can drive the cut-off close to the tabulated value.
The cut-off as a first-order signature
Before considering any time dependence, we note the simplest observational consequence of the preceding sections. Radio emission generated at the surface of a planet cannot escape below the local plasma frequency: it is reflected back towards the surface. Earth is a familiar example - short-wave broadcasting relies on exactly this reflection. The spectrum of any surface-origin/surface-dominated BRaT must therefore turn over at low frequencies, at a characteristic frequency that equation (11) allows us to estimate in advance from the stellar XUV flux, for a broadly terrestrial atmosphere: roughly 10–40 MHz for the quiescent targets of Table 2. Seen from a distance, the continuum is progressively suppressed over a band comparable in width to the cut-off frequency itself; for an Earth-like ionosphere, the disc-integrated model of § 5.2 places the decline between roughly 16 and 5 MHz (Figure 1).
Figure 1. The phase-independent turnover. The fraction of surface-origin continuum emission that escapes to a distant observer is shown as a function of frequency for two targets of Table 2: the Sun–Earth case ( MHz) and a planet in the habitable zone of AD Leo ( MHz). The lines show the orbit-averaged escape fraction of the fiducial disc-integrated model developed in § 5.2 (equation 17); the shaded bands show the range covered over the orbit, from the transit (nightside) spectrum on the left to the eclipse (dayside) spectrum on the right. The turnover moves within the band but is present at every phase. The change of slope near the half-escape level is a consequence of the sharp per-element step adopted in the model, and would be smoothed in a real ionosphere (see Figure 3).
This turnover is present at every orbital phase and for every viewing geometry, including the face-on orbits for which the phase modulation of § 5 vanishes. Its detection requires only a time-averaged, multi-band spectrum of a known BRaT, with no knowledge of the orbital ephemeris. A low-frequency turnover is not unique to this mechanism (§ 5.5), so it is a necessary rather than a sufficient condition for a surface origin. It is nevertheless an interesting test: continuum emission that persists well below the predicted plasma frequency would be very difficult to reconcile with a surface origin, and would point instead to a source above the ionosphere. The phase dependence developed in § 5 refines this basic signature into a more discriminating one.
Orbital-phase modulation as a diagnostic
Section 4 treated the ionospheric cut-off as a single number for each planet, set by the quiescent XUV flux of the host star and modified by flaring activity. In reality, the cut-off is not static in space or time. The dayside of the planet is illuminated by its host star; the nightside is not, and the local cut-off at any given location varies systematically with local time. Crucially, however, the resulting day–night pattern is anchored to the substellar point, not to the rotating surface: the surface rotates through a pattern that stays fixed relative to the star (on a synchronously rotating planet, by contrast, the pattern is fixed to the surface itself, and the surface does not rotate through it; see § 5.1). What a distant observer can detect is therefore not the rotation of the planet but the changing viewing geometry of this star-anchored pattern, which repeats on the orbital period. Throughout this section we assume leakage that is approximately steady in time and broadly distributed in longitude, and an ionosphere controlled primarily by stellar illumination. Departures from these assumptions – localised or beamed transmitters, surface geography, magnetically organised ionospheric structure, and atmospheric response lags – can introduce additional rotation-period signals, and are discussed in § 5.3 and § 7.3. Figure 2 presents the geometry of the star-planet-observer system. This section develops that observation into a candidate diagnostic for distinguishing artificial radio emission from natural sources.
Figure 2. Viewing geometry of the phase-locked cut-off. Panel (a): the orbit viewed from above the orbital plane, with the observer towards the bottom. At secondary eclipse () the dayside hemisphere faces the observer; at transit () the nightside faces the observer; quadrature () presents half of each. Panel (b): the visible disc as seen by the observer at the same three phases, shaded by the local cut-off (equation (14)). The greyscale represents the local ionospheric cut-off, not reflected-light brightness. The substellar point (S) sits at the disc centre at eclipse but is projected onto the limb at quadrature, where the terminator (dashed) crosses the disc centre; at transit, the disc sits uniformly at the nightside floor (the uniform shade reflects the single-valued floor , which subsumes dusk-to-dawn structure into one number; § 5.3 relaxes this assumption). The visible disc thus spans the full range of local cut-offs at quadrature, but a single value at transit – which is why the disc-integrated turnover is broadest at quadrature and narrowest at transit (though not sharp: once the dependence of the escape threshold on the emission angle towards the observer is included, § 5.2, elements near the limb escape only at higher frequencies at every phase).
The local cut-off varies through the day
At any given longitude on a rotating planet, the angle between the local vertical and the direction to the host star — the stellar zenith angle — varies sinusoidally with rotation,
where is the latitude of the observed longitude, is the stellar declination set by the planet’s axial tilt, is the apparent angular rate of the star in the planet’s sky (the solar-day frequency: for a prograde circular orbit, , where is the inertial rotation rate); the corresponding solar-day period, , coincides with the rotation period only when , and lengthens without bound as synchronous rotation is approached, and is the time of local noon. At the star is directly overhead; at the star is on the horizon; on the nightside formally goes negative.
Through the Chapman scaling of equation (9), the local plasma cut-off would naively follow . The Chapman expression assumes photochemical equilibrium between the ionisation by the incident stellar XUV and recombination of the dominant ions, and it breaks down on the nightside, where the source term has switched off. There the electron density persists because recombination is not instantaneous, supplemented by plasma transport from the dayside and residual ionisation by cosmic rays and energetic particles ((Kelley 2009; Schunk & Nagy 2009)). None of these processes are captured by the Chapman expression. We follow a standard phenomenological prescription and introduce a small floor on to prevent the cut-off falling to zero on the nightside,
The value of is set empirically. For Earth’s F2 layer, the observed daytime peak plasma frequency is –13 MHz, falling to –5 MHz on the nightside (Table 1; Kelley 2009, Ch. 5), giving a measured night-to-day cut-off ratio of –0.5. Matching the upper end of this range, which corresponds to a well-developed solar-maximum nighttime ionosphere, fixes $\xi \approx 0.05 (so that ). We adopt this value as our Earth-calibrated baseline throughout; in practice, the choice of rolls together all of the non-photoionisation processes that maintain the nightside ionosphere into a single number, and a quantitatively accurate model would replace it with an explicit recombination-and-transport calculation (see also § 5.3).
With this prescription, swings between at local noon and on the nightside — a peak-to-trough modulation of just over 50 per cent at the level of a single longitude in the Earth-calibrated model used here. The fourth-root smooths the variation considerably: even when swings over a factor of twenty between noon and the terminator, varies only by a factor of two.
An important consequence follows immediately. Because equation (14) depends on position only through the stellar zenith angle,
the steady-state pattern of the local cut-off is a function of local time alone; it is stationary in the frame that co-rotates with the planet–star line. A longitudinally uniform distribution of leakage sources rotating through this stationary pattern therefore produces no modulation on the rotation period - rotating a uniform source distribution beneath a star-anchored filter changes nothing. The observable modulation instead arises as the observer’s line of sight scans across the pattern with changing orbital geometry (§ 5.2). The period and phase of this idealised ionospheric pattern are set entirely by the orbital ephemeris and the system geometry, independent of the source; this geometric phase-locking is the key point. The amplitude and detectability of the observed modulation, by contrast, do still depend on the source – on its spectrum, spatial distribution, antenna patterns, and duty cycle (§ 6 and § 7.3). Genuine rotation-period signals require structure anchored to the surface or to the planetary magnetic field: most obviously a geographically concentrated transmitter distribution, as on present-day Earth, but natural, magnetically organised ionospheric or magnetospheric structure can also rotate with the planet. Rotational signals are separable from the orbital signal by timescale, but are not by themselves evidence of technology (§ 7.3).
For tidally locked planets in the habitable zones of M-dwarfs, the substellar point is, in addition, fixed in the planet’s co-rotating frame, so the pattern is static in both frames: the illumination is time-independent at each surface location, and the equilibrium description of equation (14) then applies without the rotational lag of § 5.3 (transport, magnetically organised structure, and flare-driven variability of course remain). For Proxima b the relevant period is the 11.2-d orbital period. For transiting systems, the variation is locked to the transit and occultation geometry, which is independently determined and provides a stringent cross-check. For non-transiting planets – including Proxima b itself, for which photometric searches have found no transits ((Kipping et al. 2017)) – the orbital period and phase are known from radial velocities, but the inclination, and hence the expected modulation amplitude, remains uncertain (§ 5.2).
Disc averaging: what a distant observer measures
An external observer on Earth sees a distant exoplanet as an unresolved point source and does not measure the local cut-off discussed in § 5.1. They measure a single integrated spectrum from the whole visible hemisphere of the planet, in which the dayside (with its high cut-off) and the nightside (with its much lower cut-off) blend together, weighted by the projected area of each surface element. The apparent cut-off they extract is therefore an average over the visible disc, and it varies because the visible hemisphere scans across the star-anchored pattern as the planet moves around its orbit: the observer sees the full dayside at superior conjunction, the full nightside at inferior conjunction, and a mixture of the two in between.
The observable is developed in two steps. We first define a simple geometric proxy – the projected-area-weighted mean of the vertical-incidence cut-off over the visible disc – which admits closed-form limits and fixes the phase behaviour. We then construct the actual observable, the disc-integrated escape fraction, which additionally incorporates the dependence of the escape threshold on the emission angle towards the observer (§ 2) and an explicit model for the angular emission pattern of the sources. As we show, the second step changes the level and shape of the turnover substantially, but not its fractional modulation.
A convenient proxy for this apparent cut-off is the projected-area-weighted average of the local cut-off over the visible disc. Writing for the local surface normal, for the unit vector towards the observer, and for the orbital phase ( at superior conjunction), we define
where is given by equation (14) and the integrals run over the observer-facing hemisphere. Equation (15) should be regarded as a geometric proxy for the observable turnover frequency rather than a full radiative-transfer solution. It assumes that each surface element contributes in proportion to its projected area, and it assigns each element its vertical-incidence cut-off, ignoring the fact that the radiation reaching the observer leaves each element obliquely, at emission angle , and must therefore satisfy the higher threshold of § 2. Both omissions are repaired in the escape-fraction model below; the proxy is retained because its closed-form limits make the phase behaviour transparent.
For an edge-on (transiting) geometry, two limits bound the behaviour of equation (15), and both can be evaluated in closed form (at lower inclinations the fully illuminated and fully dark hemispheres are never presented to the observer, and these extremes are not reached; see below). When the substellar point faces the observer (the secondary-eclipse geometry of a transiting system) the directions to the star and to the observer coincide, so for every surface element the projection weight equals the cosine of the stellar zenith angle, . Writing and noting that the element of surface area on the visible hemisphere is , equation (15) with reduces to a one-dimensional integral,
where the factor is the Chapman scaling of the local cut-off and the additional factor of is the projection weight. The apparent cut-off at this phase is thus dragged only slightly below the noon value by the dimmer limb regions seen in morning and evening light.
At the opposite extreme, when the antistellar hemisphere faces the observer (the transit geometry), every element of the visible disc sits at the nightside floor of equation (14), and the average is exact and trivial: . The proxy is the dotted curve in Figure 3(a): it glides smoothly between these two limits as the planet moves around its orbit, with a fractional modulation of per cent. The three principal viewing geometries – secondary eclipse, quadrature, and transit – and the corresponding appearance of the visible disc in terms of the local cut-off are also illustrated in Figure 2.
Figure 3. Disc-integrated escape model including the observer-directed escape threshold (equation 17), for the Earth-calibrated nightside floor , zero obliquity, an edge-on orbit, and uniform intrinsic leakage. Panel (a): the half-escape frequency as a function of orbital phase ( at superior conjunction). Solid: the fiducial model (thin-shell obliquity factor with and Lambertian emission weighting ). Dashed: the plane-parallel secant law , for which the eclipse and transit values are analytically, and . Dot-dashed: isotropic emitters (). Dotted: the vertical-incidence proxy of equation (15), which spans to . Including the observer-directed threshold raises the level of by per cent and produces a flat plateau at the nightside value for , but leaves the fractional modulation essentially unchanged: 48 per cent for the fiducial and plane-parallel models, against 47 per cent for the proxy; isotropic and horizon-peaked emission give 44 and 39 per cent. Panel (b): the escape fraction of the fiducial model at eclipse, quadrature and transit, with a sharp per-element step; the horizontal line marks . At eclipse no leakage escapes below , because every element must satisfy ; the extended high-frequency tails arise from elements near the limb. The sharp onset at and the steep rise near in the eclipse curve arise from the sharp per-element step and the infinitely thin shell: the observer-directed threshold peaks at near and falls again towards the limb. In a real ionosphere these features would be rounded off by finite layer thickness, refraction, magnetoionic splitting (of order ), and collisional absorption (§ 2), with little effect on .
The observable itself is the disc-integrated escape fraction. Let be the cosine of the emission angle towards the observer. Radiation from a surface element reaches the observer only above the observer-directed threshold , where is the obliquity factor of § 2: the secant law in the plane-parallel limit, or its thin-shell regularisation. The received contribution of each element also depends on the angular emission pattern of the sources, which we describe by a weighting per unit surface area. Then
is the fraction of the disc-integrated leakage that escapes to the observer at frequency and phase , where is the per-element transmission profile, a step (or smoothed step) at the observer-directed threshold. Three emission weightings are considered. Lambertian, , is appropriate to a diffuse surface of uniform specific intensity and is adopted as the fiducial case; we stress that this is an assumption, not a generic consequence of uniformly distributed transmitters. Isotropic, , describes a uniform population of discrete transmitters each radiating isotropically into the upper hemisphere. Horizon-peaked, , represents antenna patterns directed towards the horizon, as is common for terrestrial broadcast transmitters. In full generality depends also on the local magnetic-field geometry and polarisation, , and a non-uniform intrinsic surface brightness would enter the integrand as an additional weight that cancels only in the uniform case (§ 7.3). We adopt the scalar, uniform-leakage form of equation (17) with a sharp per-element step, so that the analytic limits below can be checked exactly and the effect of obliquity isolated; magnetoionic and collisional broadening would further smooth the resulting curves. We characterise the turnover by the half-escape frequency , defined by , and define the fractional modulation as .
Two limits can again be evaluated in closed form, for the plane-parallel secant law with Lambertian weighting. At superior conjunction , so the threshold is ; with , an element escapes if , and integrating the projected-area weight gives
At transit every element sits at the floor, so the threshold is and
Comparison with the proxy limits and shows that the observer-directed threshold raises the level of by per cent at both extremes – a shift of the turnover, not merely a smoothing of its edge – whereas the fractional modulation, per cent, is almost identical to the proxy value of 47 per cent, because obliquity multiplies both extremes by similar factors. With the thin-shell regularisation of § 2 () the extremes become and , again with per cent (Figure 3a). The amplitude does depend on the assumed emission pattern: isotropic emitters give spanning – and per cent, and horizon-peaked emitters – and per cent, because patterns that favour the limb sample more oblique, higher-threshold paths. We therefore quote per cent for the fiducial model and 39–48 per cent across the emission models considered, regarding the emission weighting as an explicit model assumption to be fitted or marginalised over in any real analysis.
Two further features of Figure 3 follow directly from the geometry. First, sits on a flat plateau at its transit value for (for the fiducial Lambertian weighting; the plateau is narrower for emission patterns that favour the limb): whenever the substellar point lies more than from the observer direction, the inner of the disc – half of the projected area – is entirely on the nightside, so the half-escape frequency is pinned at the nightside value regardless of the dayside sliver at the limb. Secondly, the low-frequency onset of escape is not fixed near the floor at all phases: at superior conjunction no leakage escapes below , since every element must satisfy , whereas at transit escape begins at from the disc centre. It is the half-escape frequency, rather than the onset, that we use to characterise the migrating turnover.
The shape of the transition also varies with phase in a predictable way. Measured between the 10 and 90 per cent escape levels, the fiducial transition spans , and at eclipse, quadrature and transit respectively: it is broadest near quadrature, where the visible disc spans the full range of local cut-offs, and narrowest at the transit geometry, where the disc sits uniformly at the nightside floor (Figure 2) – though not sharp, since elements near the limb escape only above , and every phase carries a high-frequency tail contributed by the limb. This phase-dependent width is itself part of the template, and provides an additional consistency check beyond the migration of alone. The family of curves is also significantly asymmetric: the quadrature spectrum resembles the transit spectrum much more than the eclipse spectrum (, and respectively). This is because the substellar region is projected onto the limb while the more highly weighted terminator region occupies the disc centre (Figure 2, middle panel of b).
The vertical-incidence proxy of equation (15) therefore underestimates the observable by per cent at all phases and is not itself the template; it is retained only as the geometric skeleton whose closed-form limits explain the phase behaviour. Figure 3(a) shows itself, and it is this curve – computed with an explicitly stated obliquity factor and emission weighting – that constitutes the template light curve used in § 6.
The fiducial geometry of Figure 3 assumes an edge-on view, in which the substellar and antistellar hemispheres each move fully into the line of sight. For lower orbital inclinations, the observer never sees the full day–night contrast: the modulation amplitude is reduced, vanishing entirely in the face-on limit of this idealised circular, zero-obliquity, axisymmetric model (eccentricity, obliquity and seasonal effects, magnetic asymmetries, or time-variable stellar activity could still induce orbital variability, though not with the specific phase relationship predicted here). This geometric dependence favours an observational programme (see § 6) that targets close to edge-on transiting systems that present the maximum modulation amplitude and the best-determined ephemerides. For a transiting system the planet is itself occulted at exact superior conjunction, so the full-dayside spectrum of Figure 3 is sampled immediately before and after secondary eclipse rather than at exactly; the plateau around transit is unaffected, since the planet is then in front of the star.
Finally, we note what survives when the phase information is discarded altogether. Averaged uniformly over the orbit, the fiducial escape fraction has a half-escape frequency of and rises from 10 to 90 per cent between and , with no escape below the nightside floor, ; the orbit-averaged half-escape frequency is and for isotropic and horizon-peaked emitters respectively. For MHz the fiducial continuum therefore declines between about 16 and 5 MHz. This is the phase-independent turnover of § 4.5: it is present in any time-averaged spectrum, at any inclination.
Departures from equilibrium: rotation and the dusk–dawn asymmetry
The equilibrium prescription of equation (14) assumes that each surface element adjusts instantaneously to its local ionising flux. In practice, the response takes a finite time (§ 4.4): the electron density of a surface parcel obeys
where the loss term is that of equation (7) and now enters as a production floor representing plasma transport and residual nighttime ionisation: a background supply that dominates wherever photoionisation falls below it. The equilibrium solution of equation (18) recovers equation (14) exactly, including the noon normalisation. For a non-synchronous rotator, each surface parcel carries its ionisation history through the day–night cycle. Writing and for the local-time angle, with the solar-day angular frequency of § 5.1, equation (18) becomes the non-dimensional equation that we actually integrate,
with . Two points about this equation should be made explicit. First, the relevant forcing period is the solar-day period of § 5.1, not in general the inertial rotation period; the two coincide for Earth but can differ greatly for close-in, resonant, or nearly synchronous planets. Secondly, the coefficient that sets the relaxation time is not the laboratory dissociative-recombination coefficient of § 4: it is an empirical effective coefficient, chosen to represent the range of observed ionospheric relaxation times. With the laboratory value, equation (12) gives a relaxation time of seconds, which is a lower bound (§ 4.4). The observed relaxation of Earth’s ionosphere is far slower, reflecting the rate-limiting ion–neutral chemistry and transport discussed there: effective response times range from tens of minutes in the lower F region to of order one to a few hours near the F2 peak ((Schunk & Nagy 2009)). We therefore present two illustrative cases that bracket this range for a 1-d solar day: a fast response, min (), and a slow response, h (), the latter being more representative of Earth’s F2 peak. Larger values of also describe planets with shorter solar days.
Figure 4(a) shows the steady-state local cut-off around a full day, obtained by integrating equation (18) to a periodic state. For finite , the cut-off no longer drops to the floor at the dusk terminator: longitudes recently in sunlight retain near-daytime densities and decay through the night, so the floor is reached only by the longitudes that have been dark longest - the pre-dawn limb that is about to rotate back into starlight. The nightside minimum is thereby localised near dawn rather than distributed uniformly. We note that this lagged pattern is still steady in the frame co-rotating with the planet-star line, so the conclusion of § 5.1 still stands: the disc-integrated modulation remains locked to the orbital phase.
The observable consequence, shown in Figure 4(b), is a distortion of the phase curve rather than a change in its period. Panel (b) shows the half-escape frequency computed from equation (17) with the lagged local pattern and the fiducial obliquity factor and emission weighting of § 5.2. For the fast response () the fractional modulation is unchanged at per cent, but the whole curve is displaced in the direction of rotation: the maximum lags superior conjunction by , and the centre of the nightside plateau lags transit by . The slow response (), closer to Earth’s F2 peak, gives lags of and respectively, with reduced to 42 per cent. The lag thus constrains the effective atmospheric response time relative to the solar-day period. It is not a clean measurement of ion chemistry alone – transport, winds, and plasma drifts all contribute to , and additional phase shifts can be introduced by a non-uniform source distribution or by eccentric illumination – but, with those caveats, its magnitude would
Figure 4. Time-dependent generalisation of the cut-off model, from the non-dimensional form of equation (18) (§ 5.3), for a solar day d and prograde rotation. Three cases are shown: the equilibrium limit (solid), which also describes synchronously rotating planets; a fast response, ( min; dashed); and a slow response, ( h; dot–dashed), closer to the behaviour of Earth’s F2 peak. Panel (a): the steady-state local cut-off at the equator as a function of local time. A finite response time carries near-daytime densities past dusk, so that the nightside minimum is reached only just before dawn. Panel (b): the corresponding half-escape frequency as a function of orbital phase, computed with the fiducial obliquity factor and Lambertian emission weighting of Figure 3. The equilibrium curve is symmetric about transit. A finite response displaces the curve in the direction of rotation: the maximum lags superior conjunction by (fast) and 0.07 (slow), and the centre of the nightside plateau lags transit by and 0.15 respectively. The fractional modulation is 48 per cent for the equilibrium and fast-response cases and 42 per cent for the slow response. In all cases the period of the modulation remains the orbital period.
bound the atmospheric response timescale independently of the flare response of § 4.4, and its sign carries information about the sense of rotation. For synchronously rotating planets, no surface parcel crosses the terminator: in this idealised circular, zero-obliquity model the forcing is time-independent, the rotational lag vanishes, and the phase curve is symmetric about the transit phase.
Flares as a competing source of time variability
The phase-locked signal described above is not the only source of time variability in the cut-off frequency. As quantified in § 4.4, stellar flares from M-dwarf hosts can transiently enhance the XUV flux at the planet and drive the cut-off upwards on timescales of minutes to hours, although the scaling limits even a 100× XUV enhancement to a factor of in , and bursts shorter than the atmospheric response time (§ 4.4) produce only a partial response.
The two sources of variability can be separated, but not trivially. In the time-series analysis described in § 6.3, flare-driven enhancements appear as aperiodic positive excursions in that are uncorrelated with the planet’s orbital period. Standard periodograms, including Lomb–Scargle, are nevertheless sensitive to large non-periodic excursions: flares raise the effective noise floor, can generate aliases through irregular sampling, and can bias the recovered amplitude and phase of the orbital signal. In practice, flare intervals should be identified through simultaneous X-ray or optical monitoring of the host star and then masked, explicitly modelled, or down-weighted, rather than assumed harmless.
Confounding sources
Several natural processes produce low-frequency turnovers in radio spectra, and several are also periodically variable. Any claim of a technological origin for a phase-locked turnover must therefore survive comparison with each of the following.
Stellar magnetic activity. Active stars display rotationally modulated radio emission tied to the stellar magnetic geometry (Güdel 2002). This variability, however, is locked to the stellar rotation period, not the planetary orbital period, and is accompanied by correlated activity indicators (X-ray, optical flaring, chromospheric emission) that can be monitored simultaneously. Such multiwavelength indicators are valuable but not definitive discriminants: the radio–X-ray relation carries substantial scatter and behaves differently for coherent and incoherent emission. The cleanest separation is by period, with near-commensurate stellar rotation and planetary orbital periods remaining the most difficult case.
Auroral cyclotron-maser instability (CMI) emission. Planetary magnetospheres are genuine periodic radio sources: Jupiter’s decametric emission is strongly modulated at the planetary rotation period (Zarka 1998), and analogous emission from exoplanets, modulated at rotational or orbital timescales, is an active area of research. CMI emission, however, differs from an ionospheric turnover in almost every other respect: it is generated at the local electron gyrofrequency rather than the plasma frequency, is strongly circularly polarised (up to per cent), and is typically bursty and narrowly beamed. CMI can nevertheless produce drifting spectral structures, frequency-dependent visibility, and broad time-averaged emission, and propagation effects or unresolved averaging can reduce the observed circular polarisation, so no single CMI characteristic is decisive on its own. The fundamental distinction is that CMI generates radiation, with spectral and temporal structure tied to the emission process itself, whereas the effect proposed here is a filter imposed on pre-existing broadband radiation, with structure tied geometrically to the planet’s day–night pattern.
Star–planet interaction (SPI). Magnetic SPI can produce radio emission modulated at the planetary orbital period or the star–planet synodic (beat) period (Zarka 2007). Where SPI emission is CMI-driven it shares the polarisation and burstiness discriminants above, but not every proposed SPI signal need be CMI-dominated: incoherent, gyrosynchrotron or other magnetic emission may also contribute. SPI is in this sense the most demanding natural contaminant, because orbital phase coherence alone cannot distinguish it from the proposed ionospheric signature. The discrimination must instead rest on spectral character (a drifting turnover frequency rather than a flux modulation), polarisation, and the location of the emission, which for SPI lies in the stellar corona rather than at the planetary position—a difference that future high-resolution astrometry can, in principle, test.
Free–free absorption in ionised winds. A dense stellar wind imprints a low-frequency free–free turnover on any embedded source (Wright & Barlow 1975). The turnover frequency, however, evolves on wind timescales and is not phase-locked to the planetary ephemeris; the discriminant here is spectral shape and phase behaviour, not periodicity alone.
Gyrosynchrotron emission from stellar coronae. Coronal gyrosynchrotron is an intrinsically generated, broad, magnetically controlled stellar component whose self-absorption turnover varies with the source opacity and the energetic-electron population (Dulk 1985). It can be modulated as active coronal structures rotate through view, and in close interacting systems may even display orbital modulation, so it is not necessarily aperiodic on planetary timescales. The distinction is again generative versus filtering: a gyrosynchrotron turnover tracks the evolving stellar source itself, whereas the ionospheric turnover is an attenuation imposed on a pre-existing planetary-surface spectrum, phase-locked to the planet’s day–night geometry and positionally coincident with the planet.
Circumplanetary plasma tori and intervening structures. Absorption by a plasma torus (an Io-like configuration; e.g. Kaiser et al. 1996), could, in principle, be modulated on the orbital period and is arguably the most awkward confounder. A torus can impose not only broadband absorption but also its own plasma-frequency cut-off, free–free absorption, refraction, and scattering, with mode-dependent propagation; and its electron density is spatially structured, as radio-occultation measurements of the Io torus demonstrate (Phipps et al. 2018). Torus effects, however, track the torus geometry and density structure, and primarily modulate the amplitude of the received flux across a broad band, whereas the ionospheric diagnostic predicts a shift of the turnover frequency itself, following the specific law of equation (14) with its maximum locked to the substellar phase.
We therefore do not claim that periodic low-frequency variability is, by itself, unambiguous evidence of technology. It is not: the literature on planetary and stellar radio emission provides many natural examples of such variability. The value of the ionospheric diagnostic lies instead in the particular observational pattern expected for planetary-surface leakage: (i) a smooth spectral turnover rather than a beamed burst; (ii) weak polarisation, in contrast to CMI processes; (iii) a characteristic frequency, not merely an amplitude, that varies with phase; (iv) modulation following the Chapman scaling, with as the principal frequency scale and a small set of geometric and atmospheric nuisance parameters; and (v) phase-locking to an independently determined orbital ephemeris, with the maximum cut-off occurring at the substellar phase or lagging it by a predictable amount set by the atmospheric response (§ 5.3). Each of these properties has natural analogues when considered in isolation; their combination, to the best of our knowledge, does not. A low-frequency turnover at the predicted frequency is thus the basic signature of a surface-origin BRaT (§ 4.5), but it is not by itself a sufficient one. It is the orbital-phase behaviour of that turnover which, combined with the associated broadband continuum detection, would substantially strengthen a technological interpretation.
Detection prospects
BRaTs are expected to be faint Jy and sub-Jy radio sources (Garrett 2026). Detecting an ionospheric cut-off is therefore considerably more challenging than simply detecting the continuum emission itself. In this section, we outline a possible long-term detection strategy and discuss the main observational challenges.
Frequency range
The most efficient frequencies for the initial detection of BRaTs are likely to lie above GHz, where the sky noise is relatively low, the planetary ionospheres are essentially transparent, and the ionospheric cut-off effects are negligible. At these frequencies, facilities such as SKA-Mid and the ngVLA are expected to provide Jy and sub-Jy continuum sensitivities over broad instantaneous bandwidths (Braun et al. 2019; Selina et al. 2018), making them natural instruments for the initial discovery of candidate BRaT systems.
A candidate identified at GHz frequencies would naturally be followed up at lower frequencies, approaching the ionospheric cut-off values predicted for exoplanets orbiting Sun-like stars and M-dwarfs. From Table 2, the expected cut-offs for most systems lie in the 5–30 MHz range, rising to 40–120 MHz only for planets around active M-dwarfs (§ 4). The first objective of such follow-up is simply to establish whether the continuum turns over near the predicted frequency (§ 4.5); this needs only a time-averaged, multi-band spectrum. The phase-resolved measurements of § 6.3 are a second, more demanding step. The principal range of interest, 5–50 MHz, is therefore only partly accessible to planned large facilities such as SKA-Low, whose nominal frequency coverage is 50–350 MHz. Ironically, the lowest frequencies are inaccessible from the ground due to the same phenomenon considered here: Earth’s ionosphere imposes a cut-off near 10 MHz and severely compromises calibration at frequencies below a few tens of MHz, where phase coherence times become very short. Facilities such as LOFAR and the LWA can observe below 50 MHz (the MWA band, by contrast, begins near 70 MHz), although with less sensitivity than SKA-Low will provide within its own band. Nevertheless, detections of bright sources, including pulsars (Pilia et al. 2016) at frequencies as low as 15 MHz are encouraging.
Sensitivity and cadence
At the lower edge of the SKA-Low band (50–60 MHz), conditions are far harsher than at the band centre. The originally projected full-station thermal sensitivity is 160 Jy beam in 1 hr (Braun et al. 2019), driven primarily by the steeply rising Galactic sky temperature, and the current SKAO performance estimates additionally quote a classical confusion limit of 680 Jy beam at these frequencies (SKA Observatory [30]). Extracting weak signals at the lowest frequencies will therefore require the longest-baseline configurations, positional prior knowledge of the target, and careful subtraction of confusing sources. The orbital-phase modulation of § 5 is therefore unlikely to be detectable in individual observations, except perhaps for the nearest and brightest BRaT candidates. The observational objective would be to recover a weak, phase-coherent modulation by combining data over many orbital cycles. For an especially compelling BRaT candidate, this could justify a dedicated low-frequency campaign with a total integration time of many weeks, perhaps months. The star-anchored nature of the pattern (§ 5.1) simplifies the observing cadence considerably. Because the apparent spectrum varies only slowly with orbital phase, individual integrations lasting a substantial fraction of an orbital phase bin can be accumulated with little smearing: a one-day integration on a ten-day orbit averages over , which for a near-sinusoidal modulation costs only per cent of the amplitude. The tolerable integration length should be determined from the adopted phase template; there is, in particular, no need to resolve the planet’s rotation. Conversely, integrations spanning one or more rotation periods deliberately average out any secondary modulation arising from a very uneven geographic distribution of surface transmitters (§ 7.3). We stress that this isolates the geometrically determined component of the modulation only under the explicit assumption that the leakage is sufficiently widespread in longitude and statistically stationary over the campaign; the recovered spectrum always retains its dependence on the transmitters’ luminosity, spectrum, and duty cycle.
Phase-folded detection
The appropriate search strategy is a matched spectral–phase template search, analogous in a general sense to periodic-signal searches in transit photometry (Kovács et al. 2002) and pulsar astronomy. Rather than attempting to measure a cut-off frequency in each observation, the data are folded on a trial orbital period and phase.
For a trial period and reference epoch , each observation is assigned an orbital phase,
Spectra obtained at similar orbital phases are then combined. If the assumed period is incorrect, the modulation is smeared out, and the folded spectrum approaches the time-averaged continuum. If the correct period is used, the ionospheric modulation adds coherently while random noise averages down. A genuine signal will manifest as a single peak centred on the phase at which the substellar hemisphere faces the observer – or displaced from it by the lag expected for a finite atmospheric response (§ 5.3), which forms part of the template rather than a discrepancy – with a fractional modulation per cent in the fiducial model (39–48 per cent across the emission models of § 5.2). The light curve remains within the upper 20 per cent of its range for only , and sits on its nightside plateau for around transit, reflecting the rapid hand-over from dayside to nightside dominance near quadrature (§ 5.2). Quantitative agreement between the observed light curve and a Chapman-model template with as the principal frequency scale – together with a small set of geometric and atmospheric nuisance parameters (the inclination, the night-side floor , the response parameter , the emission weighting and obliquity factor of § 5.2, and the transmission-profile width) – provides a constrained, predictive test that few astrophysical alternatives can pass.
Panel (a) of Figure 3 is, in effect, the template light curve against which folded data would be compared. The search is therefore fundamentally a model-comparison problem: one tests whether the phase-dependent escape-fraction model of equation (17), with the Chapman template of § 5 and as its principal frequency scale, provides a significantly better description of the accumulated data than a static spectrum. Where the orbital period is already known from transit photometry, radial-velocity, or astrometric observations, as it might well be for any credible BRaT candidate, both the expected period and phase are fixed in advance, which greatly reduces the trials factor. The analysis nevertheless remains a constrained template-fitting or model-comparison problem rather than a literal single-hypothesis test, since the atmospheric, geometric, continuum, and calibration parameters are still fitted. The signature is thus not a single spectral measurement but a coherent time-domain signal recovered through long-term integration. At frequencies where the array is not confusion-limited (§ 6), coherent integration times of several months could, in principle, approach Jy sensitivity levels, making outstanding BRaT candidates possible targets.
Verification
A convincing detection must satisfy several independent criteria, which together define how a false positive would be recognised and rejected:
(i) Phase coherence. The folded signal must strengthen as additional data are added, remain stable across independent observing campaigns, and lose coherence, or decrease substantially in significance, when the data are folded at incorrect trial periods (aliases and the observing-window function can leave residual peaks, so the signal need not vanish entirely).
(ii) Ephemeris agreement. The recovered period and phase must match the independently determined orbital ephemeris, with the maximum apparent cut-off occurring at superior conjunction in the instantaneous-equilibrium model, or displaced from it by the phase lag expected for a finite atmospheric response (§ 5.3); the comparison must be made against the appropriate equilibrium or lagged template. For a transiting system the planet is occulted at exact superior conjunction, so the full-dayside spectrum is sampled immediately before and after secondary eclipse.
(iii) Spectral localisation. The variability must be confined to frequencies near the spectral turnover, with the continuum well above the cut-off remaining steady in phase.
(iv) Weak polarisation. The modulated component should show the weak polarisation expected of incoherent broadband leakage, in contrast to the strong circular polarisation of auroral CMI emission (§ 5.5).
(v) Astrometric association. The emission must be positionally co-incident with the planetary system, sharing its proper motion and parallax.
(vi) Independence from stellar activity. Simultaneous X-ray and optical monitoring of the host star must show no correlation between the phase-locked modulation and stellar flaring. A recovered period distinct from the stellar rotation period is a strong discriminant where available, but not a necessary condition: if the two periods are equal, nearly equal, or commensurate, discrimination must rely instead on spectral morphology, multi-wavelength activity indicators, polarisation, and comparison with stellar-activity and star–planet-interaction models (§ 5.5).
(vii) Full template agreement. Beyond phase coherence at the correct period, the signal must match the complete predicted spectral–phase template: the migration of with phase, the phase-dependent turnover width (§ 5.2), and repeatability of the pattern across independent frequency sub-bands.
(viii) Terrestrial-interference rejection. Radio-frequency interference must be excluded explicitly: through simultaneous off-source reference beams, interferometric localisation of the signal to the target position, repetition across epochs and (where possible) across geographically separated stations, and standard instrumental and environmental checks.
No single criterion would be decisive on its own. However, a candidate satisfying all eight criteria would be difficult to reconcile with the main false-positive classes discussed above: terrestrial interference, stellar activity, auroral CMI emission, or unrelated background sources. In particular, a weakly polarised spectral turnover that is astrometrically associated with the planetary system and phase-locked to the independently known orbital ephemeris would elevate the source to a high-priority technosignature candidate warranting sustained follow-up.
The main practical limitations are access to the appropriate frequency range and good sensitivity. A proper exploration of the MHz regime may therefore require a major space-based or lunar-farside low-frequency radio observatory. Such facilities on the required scale are unlikely to be available in the near term, making the ionospheric cut-off a challenging diagnostic for present instruments but a potentially powerful target for future low-frequency radio astronomy.
Discussion
The cut-off as a planetary atmospheric diagnostic
Although motivated here as a technosignature discriminator, the phase-locked cut-off is, in the first instance, a measurement of planetary atmospheric physics. A detection would directly constrain the peak electron density of the ionosphere through equations (1) and ((14)). It would also constrain the day-night electron density contrast through the modulation amplitude. However, the amplitude is degenerate with the orbital inclination, the source distribution, the angular emission pattern of the leakage, and the adopted transmission profile, so the empirical floor must be inferred jointly with these parameters rather than measured directly. The observations would further constrain the nightside plasma transport and residual ionisation that physically set this floor. Finally, the effective atmospheric response time, , could be inferred from the dusk–dawn lag of the phase curve (§ 5.3). The sign of the lag additionally carries information about the sense of rotation, subject to the caveats of § 5.3. Ultraviolet transit spectroscopy, atmospheric-escape signatures, auroral radio emission, and star–planet plasma interactions have all been discussed as indirect probes of ionised upper atmospheres; the approach proposed here is unusual in offering a direct constraint on the peak ionospheric plasma frequency of a terrestrial exoplanet and on its variation in local time. The observational programme outlined in § 6 is admittedly long-term but does have scientific value beyond the field of technosignatures.
Value for SETI
The diagnostic has a property that distinguishes it from most proposed technosignatures: it depends on planetary physics, not on the behaviour of the transmitting civilisation. Most radio SETI strategies must assume something about intent—deliberate beacons, particular frequency choices, persistence, or duty cycle. The ionospheric imprint requires none of this. Any surface-generated broadband leakage, whatever its purpose, spectrum or schedule, must cross the ionosphere on its way out, and acquires the phase-locked turnover in doing so. The ionospheric transfer function itself is set by planetary physics, and its phase is compatible in advance from the system’s known geometry. Recovering it observationally does, however, still make demands of the transmitters: the leakage must contain appreciable power across the cut-off, escape upward, persist long enough for phase folding, and be reasonably stationary (§ 6 and § 7.3). What the diagnostic removes is any assumption about intent; it does not remove the need for a detectable, persistent transmitter. We note also that a photoionised ionosphere is a common consequence of any irradiated atmosphere – it is neither unique to nor guaranteed by habitability – so the cut-off converts a near-universal feature of irradiated planetary atmospheres, rather than of habitability as such, into a verification tool.
Caveats
Several simplifications in our treatment should be kept in view. The Chapman description assumes photochemical equilibrium with a single dominant ion species; real ionospheres have multiple layers, and transport becomes important at high altitudes and on the nightside, where our phenomenological floor absorbs a good deal of unmodelled physics (the time-dependent treatment of § 5.3 relaxes the equilibrium assumption but retains the single-species chemistry).
A second point concerns the geographic distribution of the leakage across the planet’s surface. Our disc-averaged cut-off behaves like a smooth, area-weighted turnover only when many longitudes contribute to the integrated spectrum. The turnover is in every case a property of the ionosphere rather than of the source. Any radiation that escapes must first pass through the ionosphere above its own emission region, whose cut-off is fixed by the local stellar illumination and hence by local time, but the source geography determines which ionosphere dominates the average. If the leakage came instead from a handful of localised transmitters, the integrated spectrum would be weighted towards the ionosphere above those sites, so that the phase and amplitude of the turnover modulation would track their particular longitudes rather than the disc as a whole. Broad longitudinal coverage must therefore remain an explicit model assumption.
It is tempting to argue that a Kardashev Type I civilisation, harnessing energy on a planetary scale, would spread its leakage across all longitudes, but planetary-scale energy use implies nothing definite about transmitter geography, spectrum, beaming, or the adoption of radio-quiet technologies. Earth at least provides an existence proof: its unintended radio emission, though clustered on land ((Saide et al. 2023)), spans the full range of longitudes, so that the disc-integrated terrestrial signal is nothing like being dominated by a single longitude. In this broad-coverage limit, the turnover frequency and any accompanying flux modulation are complementary observables with different sensitivities: both depend on the ionospheric transmission and on the source geography, but the turnover frequency responds primarily to the planetary ionosphere, while the broadband flux responds primarily to the geographic distribution and duty cycle of the emitters.
Finally, our scaling relations assume that the leakage spectrum is intrinsically broadband down to the cut-off; a civilisation whose emission is concentrated well above would carry no measurable ionospheric imprint. None of these caveats alters the qualitative conclusion, but each would need to be revisited in the interpretation of any real candidate.
Conclusions
Any radio leakage generated on the surface of a planet possessing a sufficiently ionised atmosphere – a common consequence of stellar irradiation, neither unique to nor guaranteed by habitability – must propagate through the planet’s own ionosphere before escaping into space. The ionosphere, therefore, imposes a fundamental low-frequency boundary on broadband radio technosignatures (BRaTs), analogous to the role played by the Earth’s ionosphere in terrestrial radio astronomy. In this paper, we have developed the ionospheric cut-off both as a physical constraint on the detectability of planetary-surface radio emission and as a potential technosignature in its own right.
Using a simple Chapman-layer description of photoionised planetary atmospheres, we have estimated the ionospheric cut-off frequencies expected for terrestrial planets in a range of habitable-zone environments. Because the peak electron density is set by a balance between photoionisation and recombination, the cut-off frequency depends only weakly on the incident stellar XUV flux, scaling as . Consequently, even large differences in stellar activity translate into relatively modest changes in the cut-off frequency. For planets orbiting Sun-like stars and many K-dwarfs, the cut-off is expected to remain broadly comparable to Earth’s value, while planets in the habitable zones of active M-dwarfs may exhibit cut-offs extending into the tens of MHz, potentially rising towards MHz during strong, sustained flares or in the most extreme irradiation environments.
We have also shown that the ionospheric cut-off should not be regarded as a perfectly sharp spectral boundary. Oblique propagation through the ionosphere, magnetised plasma effects, and collisional absorption all act to broaden the transition between trapped and escaping radiation. The observable signature is therefore expected to be a smooth low-frequency turnover rather than an abrupt spectral edge, with the detailed shape depending on the structure and composition of the planetary ionosphere. This turnover is the primary signature. It is present at every orbital phase and for every viewing geometry, at a characteristic frequency that can be estimated from the XUV output of the host star, and it can be sought in a time-averaged spectrum without orbital ephemeris. Because natural low-frequency turnovers exist, it is a necessary rather than a sufficient condition for a surface origin; but continuum emission persisting well below the predicted plasma frequency would be very difficult to reconcile with one.
The second main result of this paper is that the cut-off need not be static. Through the Chapman scaling, the local cut-off varies approximately as , reaching its maximum at the sub-stellar point and declining towards a nightside floor maintained by residual ionisation and plasma transport. This day–night pattern is anchored to the substellar point rather than to the rotating surface, so to a distant observer, the disc-averaged cut-off is modulated on the planet’s orbital period, sweeping between a dayside maximum at superior conjunction and the nightside floor near the transit phase. Including the dependence of the escape threshold on the emission angle towards the observer raises the absolute level of the turnover by per cent and broadens it, but leaves its fractional modulation essentially unchanged: the half-escape frequency varies by per cent over the orbit in the edge-on, Earth-calibrated fiducial model (39–48 per cent across plausible angular emission patterns of the leakage), with a phase set by planetary geometry; the amplitude additionally depends on the orbital inclination and on the spatial distribution of the sources. Departures from photochemical equilibrium displace the phase curve by an amount that constrains an effective atmospheric response timescale relative to the solar-day period, without altering the period of the modulation itself.
This behaviour suggests a new potential diagnostic for planetary-surface radio emission. A broadband radio source associated with an exoplanet that exhibits a low-frequency spectral turnover, whose characteristic frequency varies periodically with the planet’s known orbital period, would be difficult to explain through conventional astrophysical processes. While individual natural processes can reproduce a low-frequency turnover, and some can reproduce periodic variability, we are not aware of any known natural process that reproduces the specific combination of a smooth, weakly polarised turnover, a stable modulation of the turnover frequency itself following the Chapman scaling, and the geometric locking of the maximum cut-off to the planet’s substellar phase (up to a predictable atmospheric lag) (§ 5.5).
Detecting such a signal will be challenging, and we emphasise that the cut-off is best interpreted alongside other evidence: a detection satisfying the verification criteria of § 6.4 would provide strong supporting evidence, though not by itself conclusive proof of an artificial planetary-surface origin. BRaTs are expected to be intrinsically faint, and the recovery of the modulation will likely require long integrations accumulated over many orbital cycles. The cut-off can be recovered by phase-folding many thousands of hours of data at the planet’s known orbital period, in a matched spectral–phase template search analogous in spirit to box-least-squares transit searches. A two-pronged programme may therefore be envisaged: detection-optimised BRaTs continuum surveys at GHz with facilities such as SKA-Mid and the ngVLA, followed by low-frequency multi-epoch monitoring with LOFAR (10–90 MHz and 110–240 MHz; van Haarlem et al. [46]) and SKA-Low (50–350 MHz; Mellema et al. 2013), whose bands together span the relevant range, to search for the phase-locked diagnostic. Access to the lowest frequencies, below MHz, is fundamentally limited from the ground by Earth’s own ionosphere — the same physical effect that the diagnostic itself exploits. Future space-based or lunar-farside observatories (Burns 2021) would therefore provide a natural extension of the programme advocated here.
The diagnostic developed here will not be tested quickly. It requires low-frequency access and sensitivities that only future facilities can provide, together with sustained multi-epoch monitoring, ancillary host-star data, and – as the caveats above emphasise – favourable properties of the transmitting source itself. The history of SETI counsels patience and a willingness to revisit assumptions as new data emerge. If a planetary-surface technological signal is ever identified at radio wavelengths, the imprint of the planet’s own ionosphere, a filter imposed by ordinary atmospheric physics on any escaping transmission, may prove to be among the most distinctive clues to its origin.
Acknowledgements
The author is grateful to the anonymous referee for a very careful and expert review of this paper. Their detailed comments on ionospheric propagation, plasma chemistry, and the limitations of the disc-averaged model corrected several errors, sharpened the physical arguments, and substantially improved the manuscript overall. Generative artificial intelligence tools were used in the preparation of this work. (Anthropic [2]) was used in the development of the Python code for the disc-integrated escape model and the production of the figures, and with checking calculations. The concept, the scientific content, the analysis choices and the conclusions are the author’s own, and the author has reviewed and verified all AI-assisted text, code and numerical results, and takes full responsibility for the content of the paper.
Data Availability
No new observational data were generated or analysed in support of this research. The Python code used to generate Figures 1–4, including the disc-integrated escape-fraction model of § 5.2, is available from the author on request.
This paper has been typeset from a T_{\rm E}X/LATEX file prepared by the author.
References
- [2]Anthropic 2026, Claude, https://www.anthropic.com/claude
- [30]SKA Observatory 2021, SKA1 Science Performance, SKAO document SKAO-TEL-0000818, Rev. 2
- [46]van Haarlem M. P., Wise M. W., Gunst A. W., et al., 2013, Astron. Astrophys., 556, A2



